Google has pushed the release of its Gemini 3.5 Pro frontier AI model from June to July, giving the company more time to gather feedback from early testers and adjust the system. The delay comes after CEO Sundar Pichai told audiences at Google I/O in May that
For Google, July now matters more than June. The company has delayed the launch of its next frontier AI model, Gemini 3.5 Pro, moving the planned rollout to July as it spends extra time gathering feedback from early testers and tweaking the model.
Google previously said it would release Gemini 3.5 Pro in June. But in a shift learned by Business Insider, Google is now targeting a July launch. A Google spokesperson declined to comment.
The timing of the delay lands awkwardly because Google had already previewed the model at its I/O developer conference in May. Back then, CEO Sundar Pichai said it would launch “next month,” even though the model “wasn’t quite ready.”
The human story behind the calendar change is plain: real-world testing is taking longer than Google expected. The source said the company pushed the launch date back so it could spend more time gathering real-world use cases from early testers.
While Google adjusts, the competitive pressure remains relentless. The source described the pushback in a moment of intense competition among AI labs. Gemini 3 had outperformed expectations last year. but Anthropic and OpenAI are continuing to pull ahead of Google in coding—an area that has emerged as the first major enterprise use case for modern AI.
Google is also working to make Gemini 3.5 Pro stronger on tasks that unfold over longer sequences. The upcoming model is expected to be better at long-horizon tasks and at powering agents.
Feedback from earlier versions is part of the package. Google has incorporated feedback from its recent Flash 3.5 model into 3.5 Pro, the source said—confirming a theory that Business Insider floated at I/O. That feedback includes criticisms that Flash consumed tokens too quickly.
Even before a full public launch, Gemini 3.5 Pro has been available to some users on Google’s Antigravity platform and on the AI benchmarking site LMArena.
Taken together, the sequence shows how quickly the goalposts are moving. Google is not just polishing code—it is waiting for testers to produce the “real-world use cases” it wants, while rivals keep pushing their own capabilities forward.
For now, the launch window has moved to July, and with it comes a higher-stakes question: whether Gemini 3.5 Pro can close gaps in the enterprise areas where competition is already being decided.
Google Gemini 3.5 Pro Sundar Pichai artificial intelligence Anthropic OpenAI enterprise AI coding Antigravity platform LMArena token consumption Flash 3.5